How to tell whether a table is still maintained
A maintained table leaves six traces, and every one of them can be checked from a browser in under an hour without asking anybody for permission.
Six signals a table leaves when somebody still edits it
Maintenance is not a feeling about a page, it is a set of traces. A table that somebody still edits will show a recent date somewhere, will contain at least one row that was changed after the page was first published, and will have a place where its own mistakes are recorded. A table that nobody edits shows none of those and cannot fake them for long.
The six signals below are ordered by how hard they are to fake. A visible date is easy to produce and easy to fake, so it sits first. A change history that names what was removed and why is hard to produce and nearly impossible to fake convincingly, so it sits last.
Work through them in that order and stop when two consecutive signals fail. One failure is a maintenance gap; two failures in a row mean the table is describing an older version of the world, and every figure in it needs re-checking before you act on it.
- Signal one: a date that is recent and specific. Pass standard: a day, not a month, and no more than one quarter old. Re-check: look for the same date repeated on every row, which is a stamp rather than an edit.
- Signal two: rows that differ in age. Pass standard: some rows are newer than others. Re-check: sort the rows by date and see whether the newest quarter is empty.
- Signal three: removals that are explained. Pass standard: at least one line says a figure was withdrawn. Re-check: compare two versions of the same page if you kept one.
- Signal four: a stated method for each figure. Pass standard: sample counts, units and a measuring rule. Re-check: try to reproduce one figure yourself.
- Signal five: contact that a reader can use. Pass standard: a place to send a correction that is not a comment thread. Re-check: send one correction and see whether it changes anything.
- Signal six: the table survives a cold read. Pass standard: an unfamiliar reader can tell what each column is evidence of. Re-check: hand one row to somebody who has never seen the site.
Date stamps, and the three ways they mislead
A date is the cheapest maintenance signal and the most deceptive. The first deception is the bulk stamp: every row carrying the same date, which usually means a template was refreshed rather than a figure re-checked. The second is the rolling date, where a page updates its own header automatically and looks alive while the content stands still.
The third deception is the honest one and still misleading. A maintainer re-read a row, found it correct, and updated the date. That is real work and it improves confidence, but it does not mean the underlying listing was re-opened, so a row can legitimately carry this month date and last year contents.
The test that cuts through all three is to compare a date against a change. Ask what else in the row moved on the same day: a figure, a note, a link. A date with no accompanying change is a signature, and signatures are cheap.
- Bulk stamp: same date on every row. Treat the table as refreshed rather than checked.
- Rolling stamp: a header that follows the calendar. Look for a fixed date inside the content instead.
- Reverified stamp: an honest re-read with no figure change. Useful, but it does not date the source.
| # | Date pattern | How it looks | What it usually means | Test to apply |
|---|---|---|---|---|
| 01 | Bulk stamp | The same date on every row of the table | A template was refreshed, not a figure re-read | Look for a row whose figures differ from the last version |
| 02 | Rolling header | A date that always matches today or this month | The page updates itself while content stands still | Find a fixed date inside the content rather than in the header |
| 03 | Honest re-read | A recent date on a row whose figures did not move | Somebody checked and the row still held up | Ask which source was re-opened on that day |
| ∑ | Only the third pattern represents work. It is still not proof that the listing was re-opened, which is why the link sample matters. |
Compare a date with a change, never with the calendar.
Row-level signals: dead links, repeated blocks, orphan notes
Rows fail in patterns, and the patterns are visible without opening every link. Open ten rows chosen at random rather than the first ten, and record how many resolve to the same item they describe. A table with a healthy maintenance habit usually fails a few; a table that has been abandoned fails most of the sample in a way that is obvious within minutes.
Repeated blocks are the second pattern. When twenty consecutive rows share the same note wording, the table has been filled from one source rather than checked from many, and the shared note is doing the work of an assumption. Orphan notes are the third: comments that reference a conversation, a photograph or an order that a new reader cannot find, left behind when the rest of the context moved on.
None of these faults makes a table worthless. They tell you which parts of it to treat as a starting point and which parts to treat as a finished answer, and that distinction is the entire purpose of the exercise.
- Dead link sample: open ten rows spread across the table, not the first ten, and count how many describe the item they reach.
- Repeated block: find any note that appears more than five times and treat it as a template rather than an observation.
- Orphan note: any row referring to a photograph, a message or an order you cannot open is background, not evidence.
- Stale price claim: any row quoting a number that no longer matches the listing is a row whose other figures deserve a second look.
Why a maintained table still fails on the day you need it
A table can be genuinely maintained and still be useless for the question you brought to it. Maintenance keeps a record honest; it does not make the record complete. Tables are usually maintained for the decisions their keepers make repeatedly, which may not include yours.
Availability is the usual gap. A table may be excellent on sizing and packing and silent on whether a seller will dispatch this week, because the keeper never needed that answer. Currency is the second gap: a row can be a year old and perfectly accurate about a model that has not changed.
The practical response is to check the table against your own decision rather than against a standard of quality. Write the one question you need answered, then look for the trace that answers it. If no trace answers it, the table has done its job by showing you where the work starts.
The check most often skipped: reading the change history
Almost nobody reads what a table removed, and removals are the most informative entries on any maintained page. A withdrawn figure says that a measurement did not repeat. A withdrawn row says that a link stopped describing an item. A changed rate says that the old rate was wrong rather than that the world moved.
Community reports in August 2026 described one platform as no longer taking new orders. This desk does not repeat that conclusion, because a report is not a check. What follows is the sequence you can run yourself, in order, and each step names the evidence to keep.
- Domain step: read the registration expiry and the certificate issue date. Evidence to keep: a dated note of both figures, not a screenshot of a page.
- Response step: fetch a page and an interface response, and read the body rather than the status. Evidence to keep: whether the response contained data or only an empty shell.
- Payment step: open the funding path and note which methods are offered today. Evidence to keep: the list as it stands, with the date you saw it.
- Order step: check whether a warehouse is still receiving and whether a balance can still be withdrawn. Evidence to keep: the reference of any movement, and the date it moved.
Running the whole checklist in one sitting
The six signals, the three date tests and the four verification steps fit into a single session if you keep to time. Give the signals twenty minutes, the link sample fifteen, the change history ten, and the verification steps the rest. Write one line per item as you go, and resist the urge to investigate anything that is merely interesting.
The output of the session is a short sheet with a verdict per signal and a date. Keep it, because the value of this exercise compounds: a table checked twice, three months apart, tells you whether the maintenance habit is holding, which no single check can show.
- Start with the newest date you can find and work backwards, because forward reading flatters the table.
- Sample the middle of the table rather than the top, because the top is where maintainers live.
- Record failures as counts, not adjectives, so next quarter comparison is possible.
- Stop after two consecutive failures and act on that finding instead of completing the list.
What to keep as evidence, and which question to ask whom
Keep four things: the date you checked, the sample you opened, the figure you could not reproduce, and the reply you received. Those four items settle most later disputes, and none of them requires anybody permission to collect.
If you are holding an order rather than reading a table, ask the desk that holds the order and ask about the order specifically: reference, last movement, next expected event. If you are reading a table, ask the maintainer about a row and ask about the method, not the conclusion. The two conversations have different answers, and mixing them is how a simple question turns into a thread.
What this note comes down to
- 01Maintenance leaves six traces, ordered from easiest to hardest to fake: a specific date, rows of differing age, explained removals, a stated method, a usable correction path, and a cold-read test.
- 02Date stamps mislead in three ways — bulk stamps, rolling headers and honest re-reads that never re-open the source — so compare each date against a change in the same row.
- 03Read what a table removed; withdrawn figures and retired rows are the most informative entries on a maintained page.
- 04Platform questions are answered by four checks — registration and certificate dates, response bodies, funding paths, and whether orders still move — not by a report repeated from elsewhere.
The data point behind this note
Public source check, KB-2026-09: all eight destination entries were re-dated to 2026-09-18 after each threshold and rate band was read again against a public notice, and every entry carries that date next to its basis rather than on a page header.
KB-2026-11· 2026-09-30 Batch ledger · Challenge a figure
Notes in the same cluster
| # | Note | What it argues | Words |
|---|---|---|---|
| 01 | What the table actually is, and what it was never meant to do | A CNFans spreadsheet is a hand-kept list of store links with working notes attached, and the useful question is never what it is but what it can be asked to prove. | 1670 |
| 02 | Reading a row in ninety seconds | Ninety seconds is enough to read one table row properly, provided you spend them on four fields in a fixed order instead of on the whole line at once. | 1425 |
| 03 | A first order, walked through step by step | A first order moves through six checkpoints, and each checkpoint has one question, one piece of evidence and one thing you should write down before moving on. | 1695 |
Related entries and gauges
- How do you check whether a table is still being maintained?
- How do you verify a platform is still operating before trusting it?
- Slip cross-check bench
Compare what the slip claims against what you measured.